The environment is more my area of interest, so I’m going to focus on that part.
Before I looked into AI’s environmental impacts, I too had thought it might be overfocusing a bit on the wrong areas, but I didn’t realize how much the order of magnitudes had changed. Before the large AI models we’re seeing now, data centers weren’t a major source of change in energy consumption. Overall power consumption in places like the US had been mostly level for the previous 10-20 years (up until 2020). But AI is not like most past datatcenter workloads, it is constantly high power usage. Especially for model training, it’s using the equipment at full utilization for almost the entire time. It’s using higher energy chips and far more chips overall. Besides training, typical datacenter workloads before high AI usage weren’t super high energy per request, but that isn’t true of AI either. The rapid increase in the energy consumption from it is what’s driving the issue
It’s causing us to delay closing of fossil fuel plants. It’s making previous declining datacenter energy stop declining and go the opposite direction and projected to increase datacenter energy usage go up by 165% by 2030
In Europe too, a data center-led surge in power demand is under way, after 15 years of decline in the power sector. Having surveyed utilities across the continent, Goldman Sachs Research found that the number of connection requests received by power distribution operators (a leading indicator of future demand) has risen exponentially over the past couple of years, mostly driven by data centers.
If we were talking about water usage of AI and someone brought up agriculture’s (especially animal agriculture) more dominant use, that would be fair to mention and talk about. But that doesn’t excuse AI’s water usage, just pose another area to also focus on
The environment is more my area of interest, so I’m going to focus on that part.
Before I looked into AI’s environmental impacts, I too had thought it might be overfocusing a bit on the wrong areas, but I didn’t realize how much the order of magnitudes had changed. Before the large AI models we’re seeing now, data centers weren’t a major source of change in energy consumption. Overall power consumption in places like the US had been mostly level for the previous 10-20 years (up until 2020). But AI is not like most past datatcenter workloads, it is constantly high power usage. Especially for model training, it’s using the equipment at full utilization for almost the entire time. It’s using higher energy chips and far more chips overall. Besides training, typical datacenter workloads before high AI usage weren’t super high energy per request, but that isn’t true of AI either. The rapid increase in the energy consumption from it is what’s driving the issue
It’s causing us to delay closing of fossil fuel plants. It’s making previous declining datacenter energy stop declining and go the opposite direction and projected to increase datacenter energy usage go up by 165% by 2030
https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030
If we were talking about water usage of AI and someone brought up agriculture’s (especially animal agriculture) more dominant use, that would be fair to mention and talk about. But that doesn’t excuse AI’s water usage, just pose another area to also focus on